SBIR-STTR Award

The use of expert system technology to provide aid to novice air force personnel in the performance of their jobs is a natural outgrowth of both current expert system technology and the need for better computer-aided job assisting and job training. Tdc has defined and developed a prototype methodology for the semiautomated acquisition of objects (i.e., lexical terms or dictionary entries) and the simultaneous, semi-automatic elucidation of rules from text software documents which are taining guides or automated training guides. These guides are written in natural language, and the development of expert training systems or expert job aiding always requires that a knowledge engineer reduce the training data into expert rules for a rule base. The tars method uses an initial reduction of natural language into an intermediate english/logic language composite, then further reduces the english/logic composite to a form very close to the final rules desired. The preliminary methodology involves a series of steps for the knowledge engineer similar to the series of steps performed in the writing of well-structured software in algorithmic languages.

Keywords:

The textual automated reduction system (TARS) is a unique naturallanguage (NL) parser which acts to reduce a stream of narrow domain text into an intermediate form, which itself is converted into a stream of single-clause simplified english language constructs connected by logical connectives. From this latter form, the clauses are converted to prototypical expert system "rules" based on a configuration file which directs the conversion. The entire system is based upon real-time feedback between all three conversion processes, resulting in a natural-language spreadsheet operation for the user that is intuitive and friendly. By avoiding exhaustive NL conversion and analysis, tars is able to perform nl stream conversions in real-time, and is unique in requiring that an englishspeaker (the user) be kept in the conversion loop as the final arbiter of semanticallyor syntactically difficult sentence conversions. Since the english-language speaker acts as the final authority on "meaning", pitfalls inherent in the absolute identification of "meaning" from the source text are avoided.

Keywords: